Query Garmin health data (e.g. heart rate, sleep, stress, body battery, steps) for one field over a date range. Use query_context instead for weather, pollen and air quality; call list_available_fields first if the exact field name is unknown.

field: a health field name, e.g. "resting_heart_rate", "hrv_last_night", "body_battery_max", "steps_series". A "_series" suffix means intraday timeseries data; a plain name means one value per day. Pass the field that matches what you want — "sleep_score" returns three related fields together (sleep_score, sleep_score_feedback, sleep_score_qualifier). date_from/date_to: ISO "YYYY-MM-DD", inclusive. resolution is accepted but ignored — the field name alone decides the shape.

Unknown or ambiguous field names return an "error", often with a "did_you_mean" list of candidates (e.g. "steps" or "stress" without suffix: choose the daily or the _series variant). A clear typo or alias is resolved automatically; _meta.field_resolved_from / field_used then show what was substituted. A field with no data in the range returns empty values, not an error. Every field carries a "unit".

Query external context data (weather, pollen, air quality) for one field over a date range. Use query_health for Garmin body data (heart rate, sleep, stress, ...) instead; call list_available_fields first if the exact field name is unknown.

field: a context field name (e.g. "temperature_max", "pollen_grass_series"), or a category name — "weather", "pollen" or "air" — to get the main fields of that category at once. A "_series" suffix means intraday timeseries data; a plain name means one value per day. date_from/date_to: ISO "YYYY-MM-DD", inclusive. resolution is accepted but ignored — the field name alone decides the shape.

Unknown or ambiguous field names return an "error", often with a "did_you_mean" list of candidates. A clear typo or alias is resolved automatically; _meta.field_resolved_from / field_used then show what was substituted. A field with no data in the range returns empty values, not an error.

Query the rolling 90-day baseline and statistical reference range (mean +/- 3 standard deviations) for one health field over a date range. Use this to see what "normal" has looked like for a field recently, or alongside query_health() to compare today's value against the person's own history rather than a fixed threshold.

Find days where one health field broke from its own recent pattern. Returns two kinds of findings: "outliers" (single days far outside the normal range) and "streaks" (five or more consecutive days consistently above or below the baseline average, even if none of them individually qualifies as an outlier -- useful for catching a slow drift, e.g. a gradually rising resting heart rate).

List days with no usable data over a date range. domain is "health", "context", or "fit" ("fit" always degrades to "domain not yet available" -- no FIT pipeline exists yet). A gap is a day that was never synced at all, or was synced but came back empty/failed.

Measure how closely two fields (health or context, any combination) move together over a date range -- e.g. does stress correlate with pollen count, or resting heart rate with sleep duration. lag_days (-7 to +7) shifts field_b relative to field_a: a positive lag compares field_a on a given day against field_b lag_days days LATER (useful for "does X today predict Y a few days from now" questions); 0 (the default) compares same-day values. Returns both Pearson (linear relationship) and Spearman (monotonic relationship, more robust to outliers) correlation coefficients, each in [-1, 1], plus how many overlapping data points they were computed from.

Query FIT activity data for a field over a date range. Not yet available (FIT pipeline is v1.8) — returns a clean "not available" result until then, never an error.

Query archived Garmin data as stored in raw/, without value extraction, for fields the processed tools cannot reduce to a daily value (lists, event logs). Interpreting the content is up to the caller, and no units are provided. Use query_health instead whenever the data you need exists there: three of these fields have a processed counterpart there for everyday questions (floors: floors_climbed, daily_weigh_ins: body_weight, intensity_minutes_data: intensity_min_moderate and intensity_min_vigorous).

field: one of the raw field names (health domain only; these are NOT listed by list_available_fields): daily_weigh_ins, menstrual_calendar_data, pregnancy_summary, lifestyle_logging_data, nutrition_daily_food_log, nutrition_daily_meals, nutrition_daily_settings, floors, intensity_minutes_data, body_battery_events, lactate_threshold, running_tolerance. date_from/date_to: ISO "YYYY-MM-DD", inclusive. domain: only "health" holds raw data.

Only the requested field is returned. A field name that is almost right is corrected automatically (_meta.field_resolved_from / field_used show the substitution); an unknown name returns an error, with did_you_mean when similar names exist. A domain other than "health" returns an error.

Result: {"health": {field: {"values": [{"date": ..., "raw": ...}], "source_resolution": "raw"}}}. A day that is not in the archive has no entry; an archived day with nothing stored for the field has "raw": null. Read-only; does not change the archive.

Read archive-state metadata: how complete and healthy the archive is, plus sync and quality logs. kind selects the artefact:

• "stats": small coverage/quality overview (total, high/standard/failed counts, date range, coverage_pct, last sync, integrity warnings). Use this for "how big/complete/healthy is my archive" questions.

• "device_table": devices seen in the archive, with date range and number of days each.

• "quality_log": per-day quality register.

• "source_api_log": one record per data fetch (time, source, validator status, endpoints fetched).

• "token_log": token events (created/invalidated/blocked/valid), never token contents.

• "capability_config": which optional Garmin endpoints are enabled (may be empty).

• "daily_logs", "fail_logs", "recent_logs": sanitized sync log lines, one entry per log file.

date_from/date_to (ISO "YYYY-MM-DD", inclusive) are needed for "quality_log", "source_api_log", "daily_logs", "fail_logs" and "recent_logs": without both dates these return an empty result (no error, no default range). They are ignored for the other kinds. The log kinds are large (a single week of daily_logs is several hundred KB), so request only the days you need and prefer "stats" for overview questions.

Result: {"data": ..., "error": None}. An unknown kind raises an error.

List the queryable fields, grouped by domain and source. Call this first when the exact field name for query_health or query_context is unknown (the fields of query_raw are not listed here; they are named in that tool's description). Omit domain for the full overview, or pass "health" or "context" to narrow it — the narrowed result is much smaller. "fit" returns an empty list (not available yet); an unknown domain returns no fields rather than an error.

Result: "domains" (the domain names), "metadata_kinds" (the archive- state kinds get_archive_metadata works with), "fields" ({domain: {source: [field, ...]}}; an empty list for "fit") and "units" ({field: unit} covering every field returned; values such as "index" or "text" mark fields without a physical unit). A field name ending in "_series" holds an intraday timeseries; any other name holds one value per day.

Manually trigger a SQLite cache sync against the archive — use this if recent archive changes (a sync just run, a backfill/recheck just completed) might not yet be reflected in query results. Runs the same sync the server already performs automatically at startup. May take a while on a large pending delta (long idle period since the last sync) — this call blocks until the sync finishes.